097: A Non-Destructive Hyperspectral Imaging Technique for Predicting Brix in Grapes

Introduction

Sugar content and fruit ripeness are crucial for grape harvesting and directly impact the quality of products such as jam, juice, and wine. Accurate and rapid assessments are essential for growers to optimize field practices. Sugar contents in grapes is usually represented by °Brix values and is measured using a refractometer. This conventional method is destructive and labor-intensive, presenting significant challenges for accurate and efficient onsite evaluations. Therefore, the objective of this study is to explore the feasibility of a novel sensing technique utilizing hyperspectral imaging for the non-destructive prediction of °Brix values in Concord grapes.

Methods

In this study, a total of 770 Concord grapes were manually picked from the research vineyards at Cornell Lake Erie Research and Extension Laboratory (CLEREL) during the harvest seasons of 2023 and 2024. The grapes were imaged in reflectance mode by a hyperspectral imaging system covering the spectral range of 400 nm to 1000 nm. Immediately following imaging, the reference °Brix values of individual berries were measured using a lab refractometer. Partial least squares regression (PLSR) models were constructed to investigate the relationship between spectral responses and °Brix values. Additionally, a wavelength selection algorithm based on the successive projection algorithm (SPA) was developed to minimize collinearity among predictor variables and identify the key wavebands for predicting °Brix values.

Results

The PLSR model developed using the full spectrum data attained an adjusted R2 of 0.93 and a ratio of prediction to deviation (RPD) of 3.85 when applied to unseen datasets. Furthermore, 9 wavelengths (400 nm, 425 nm, 547 nm, 611 nm, 665 nm, 692 nm, 736 nm, 918 nm, and 963 nm) were selected to construct the prediction model, which still achieved good generalizability with an adjusted R2 of 0.85 and a RPD of 2.62. The results highlight the potential of hyperspectral imaging as an effective non-destructive method for assessing grape quality.

Significance

Hyperspectral imaging-based °Brix prediction provides a non-destructive and high-throughput method, demonstrating its potential as a rapid in-situ sensing tool to help growers efficiently and accurately estimate the grape sugar contents and ripeness for optimized harvesting and quality management.

Authors: Jinhong Yu, Yu Jiang, Terry Bates, Chang Chen